Effects of Missing Data : Ranked Set Sampling vs Simple Random Sampling

Carlos N. Bouza Herrera — Universidad de La Habana
vendredi 9 décembre 2016

Missing data is a well-recognized problem which arises in statistical inferences and data analysis. Statistical methods are affected by the missingness of data.
We consider its effect in sampling finite populations. Generally surveyors decide to subsample the non-respondents when the response rate is lower than expected and to interview all of them is too costly. Non response constitutes an important potential source of bias.
Rubin (1987, 1976), Little and Rubin (1987) conspired that missing data problems may be solved by imputing them. They classified missing data mechanisms into three types. They work efficiently used in the presence of a Missing At Random mechanism. We consider that this mechanism is generating non response.
We present the behavior of ratio based imputation when estimating the population mean when Ranked set sampling (RSS) is used. This sampling design was first proposed by McIntyre (1952) and has performed adequately for different inferential problems.
We are going to discuss the behavior of some ratio based imputation procedures and compared the efficiency of their RSS counterparts compared with simple random sampling